feat(openai): add OpenAI MCP server synced to current Docs spec

[Copilot]

The AceDataCloud Docs dropped gpt-5.4 from the openai OpenAPI spec (canary sync test). No openai/ subdirectory existed in the monorepo, so this PR creates it from scratch, aligned to the current spec.

New service: openai/

Follows the same architecture as all other services (serp, shorturl, etc.).

Tools

Tool	Endpoint
openai_chat_completion	POST /openai/chat/completions
openai_embeddings	POST /openai/embeddings
openai_image_generate	POST /openai/images/generations
openai_responses	POST /openai/responses
openai_image_edit	POST /openai/images/edits
openai_list_models	informational — lists all models by category

Model coverage (from spec)

• Chat completions: gpt-5.5, gpt-5.5-pro, gpt-5.2, gpt-5.1, gpt-5, gpt-4o, gpt-4.1, o1–o4-mini families — no gpt-5.4
• Embeddings: text-embedding-3-small/large, text-embedding-ada-002
• Images: dall-e-3, gpt-image-1/1.5/2, nano-banana/2/pro
• Responses: full extended model list matching spec — no gpt-5.4

Structure

openai/
├── core/          # config, client, server, exceptions, oauth
├── tools/         # openai_tools.py + info_tools.py
├── prompts/
├── tests/         # unit tests for client + config; integration stubs
├── main.py
├── pyproject.toml
└── Dockerfile / deploy / vscode / ...